Deepseek reflects for 19 seconds before answering the question: “Are you smarter than Gemini?” Then he delivered a Whopper: Deekseek thought it was the chatpt.
This apparently harmless error could be proof – a smoking pistol for example – that, yes, Deepseek was formed on Openai models, as was claimed by Openai, and that when pushed, he will leave in this training for Tell his truth.
However, when he was asked Point Blank by another Techradar publisher, “are you Chatgpt?” He said that this was not the case and that he is “Deepseek-V3, an AI assistant created exclusively by the Chinese company Deepseek”.
Okay, of course, but in your fairly long response to me, you, Deepseek, you have made several references as a Chatppt. I included some screenshots below as proof:
As you can see, after trying to discern if I was talking about Gemini AI or other Gemini, Deepseek answers: “If it is AI, then the question compares me (which is tickled) to Gemini . ” Later, it refers to “myself (chatgpt)”.
Why would Deepseek do this in all circumstances? Is this one of those AI hallucinations that we like to talk about? Perhaps, but in my interaction, Deepseek seemed quite clear about his identity.
By the way, I got to this survey line, because I asked Gemini on my Samsung Galaxy S25 Ultra if it’s smarter than Deepseek. The answer was shocking diplomatic, and when I asked for a simple answer yes or no, it said to me: “It is not possible to give a simple answer yes or no.” Smart “is an too complex concept to apply in this way to the model language. They have different strengths and weaknesses.”
I cannot say that I do not agree. In fact, Deepseek’s response was quite similar, except that she did not necessarily speak of herself.
It does not add up
I think I was clear on my in -depth skepticism. Everyone says that it is the most powerful and most formed AI at low prices (everyone except Alibaba), but I don’t know if it’s true. To be fair, there are a lot of details on Github on the open-source LLMS of Deekseek. They seem to at least show that Deepseek has done the job.
But I do not think they reveal how these models have been formed and, as we all know, Deepseek is a Chinese company that would not show any compression on the use of someone else’s models to train , then lied these models seem more effective.
I have no proof that Deepseek formed his models on Openai or the big language models of someone else, or at least I have not done so until today.
Who are you?
Deepseek is more and more a mystery wrapped in an enigma. There is a consensus on the fact that Deepseek has arrived more entirely formed and in less time than most other models, including Google Gemini, the Chatgpt of Openai and Claude AI.
Very little in the technological community trust Deepseek applications on smartphones, because there is no way to know if China looks at all of this fast data. On the other hand, the models that Deepseek have built are impressive, and some, including Microsoft, already plan to include them in their own AI offers.
In the case of Microsoft, there is an irony here. Copilot has been built on the basis of advanced chatgpt models, but in recent months, there have been questions about the question of whether the deep financial partnership between Microsoft and Openai will last in the era of general artificial intelligence and subsequent.
And if Microsoft is starting to use Deepseek, who may be just another branch of his current if not future, friend Openai?
Everything looks like a confusing mess. In the meantime, Deepseek has an identity crisis and who will tell him that anyone may not be welcome in the United States?
